Abstract

Background: BCG vaccination has been reported to reduce all-cause infant mortality beyond its effect on tuberculosis, an example of what are termed non-specific effects of vaccination. A review commissioned by the World Health Organization Strategic Advisory Group of Experts assembled the randomised and observational evidence and called for further trials. Methods: Systematic reviews and randomised trials reporting all-cause mortality by BCG receipt in children under five were eligible. Estimates were compared across designs by ratios of relative risks with z-tests on the difference in log estimates. Absolute risk reductions and numbers needed to vaccinate were derived across plausible baseline mortality rates. Analyses were performed in Python 3. Results: Across five randomised trials the pooled relative risk for all-cause mortality was 0.70 (95% CI 0.49-1.01), corresponding to z = –1.93 and p = 0.053—an estimated 30% relative reduction whose interval crosses the null in the second decimal place. Across 13 observational studies the pooled estimate was 0.47, approximately 1.5-fold stronger, which is the ordering that healthy-vaccinee bias would produce. In two trials restricted to low birthweight infants the estimate was 0.52 (0.33-0.82), the only pooled estimate examined that excludes the null, although it did not differ significantly from the all-trials estimate (ratio 0.74, 95% CI 0.42-1.33, p = 0.316). Because relative effects translate into absolute benefit in proportion to baseline risk, the number needed to vaccinate falls from approximately 333 where baseline mortality is 1% to approximately 21 among low birthweight infants where it approaches 10%. Conclusions: The randomised evidence is consistent with a clinically important reduction in all cause infant mortality and stops just short of conventional statistical significance. The usual remedy—more placebo-controlled trials—is not available, because BCG has been part of routine immunisation since 1974 and withholding it is not ethically permissible. The limitation is structural rather than a matter of insufficient effort. Trials randomising the timing of administration rather than its receipt remain available and are the design capable of resolving the question.

Keywords: BCG, Non-specific effects of vaccines, All-cause mortality, Healthy-vaccinee bias, Low birthweight, Trial design